search

ROCm / ROCmValidationSuite

The ROCm Validation Suite is a system administrator’s and cluster manager's tool for detecting and troubleshooting common problems affecting AMD GPU(s) running in a high-performance computing environment, enabled using the ROCm software stack on a compatible platform.
MIT License
56 stars 36 forks source link

Gemm self & accuracy check feature #766

Closed jkottiku closed 4 days ago

jkottiku commented 5 days ago
  1. Validate gemm output for consistency/repeatability by self-check (current vs previous output).
  2. Validate gemm output for accuracy by accuracy-check (device (GPU) output vs host (CPU) output).